Borisevelson_3By Boris Evelson

While I echo my colleagues' earlier comments on the Microsoft/FAST Search transaction, I also give Microsoft thumbs up for being the first of the major BI vendors to embrace alternative DBMS for BI. For a while now I've been predicting that alternative DBMS for BI will gain continually increasing momentum for the following reasons:

  • Traditional relational databases were designed from the ground up for transaction processing, not BI. Only in the last decade have they even begun to accommodate BI-style queries, and still play a constant balancing act between OLTP and OLAP optimization. Columnar databases, such as Vertica, Sybase IQ, KX, ParAccel, SAND Technology, InfoBright, are specifically designed and optimized for nothing but OLAP query processing. Their schemas are also much more flexible since it's as easy to drop, add, or update a column in a columnar database as it is to insert, change, or delete a row in a relational database.
  • A similar argument points to the fact that relational databases are pushing their limits in their capabilities of storing and processing unstructured content. Search, storage, and processing are less of an issue here; instead the inherent inflexibility and rigidity of relational schemas don't lend themselves naturally to unstructured content processing. An I&KM professional typically needs to know the kinds of questions (queries) that will be asked and what types of future analysis will be required to model data for structured BI. But unstructured content in BI is unpredictable by nature, having no room for predefined schemas. Index-based DBMSes such as those from FAST Search and Endeca, on the other hand, are nothing but indexes with embedded data. They require no schemas, are infinitely more flexible than relational databases, and can handle structured data just as naturally as they handle unstructured content.

I am sincerely hoping that Microsoft/FAST Search transaction is yet another confirmation of this trend. Among other large BI vendors, SAP has already bought into the trend with SAP BI Accelerator appliance which is based on a columnar-like search index.

What's next? SAP, Oracle, or IBM may acquire Endeca — major FAST Search competitor. Sybase IQ, the major columnar DBMS player, must also be a target on Oracle or IBM's list.